This article introduces an innovative classification methodology for identifying nanowires within scanning electron microscope images. Our approach employs advanced image manipulation techniques in conjunction with machine learning-based recognition algorithms. The effectiveness of our proposed method is demonstrated through its application to the categorization of scanning electron microscopy images depicting nanowires arrays. The method's capability to isolate and distinguish individual nanowires within an array is the primary factor in the observed accuracy. The foundational data set for model training comprises scanning electron microscopy images featuring 240 III-V nanowire arrays grown with metal organic chemical vapor deposition on silicon substrates. Each of these arrays consists of 66 nanowires. The results underscore the model's proficiency in discerning distinct wire configurations and detecting parasitic crystals. Our approach yields an average F1 score of 0.91, indicating high precision and recall. Such a high level of performance and accuracy of the ML methods demonstrate the viability of our technique for not only academic but also practical commercial implementation and usage.

Feature attribution explains Artificial Intelligence (AI) at the instance level by providing importance scores of input features’ contributions to model prediction. Integrated Gradients (IG) is the prevalent path attribution method for deep neural networks, involving the integration of gradients along a path from the explained input (explicand) to a counterfactual instance (baseline). Nonetheless, current IG variants exclusively consider the gradient of explicand’s output, but we have discovered that the gradient of the counterfactual output also exerts a substantial influence on feature attribution. To achieve this, we propose Iterative Gradient path Integrated Gradients (IG2), considering both two gradients. IG2 incorporates the counterfactual gradient iteratively into the integration path, consequently obtaining a novel path (GradPath) and a novel baseline (GradCF). These two novel IG components effectively address the issues of attribution noise and arbitrary baseline choice in earlier IG methods. IG2, as a path method, stratifies many desirable axioms, which are theoretically justified in the paper. Experimental results on XAI benchmark, ImageNet, MNIST, TREC questions answering, wafer-map failure patterns, and CelebA face attributes validate that IG2 delivers superior feature attributions compared to the state-of-the-art techniques.

Our paper compares the merits of using both adaptive load resistance and adaptive frequency for impedance compensation of large spatial variations in capacitive power transfer (CPT) systems. We show through an equivalent circuit model how changes to load impedance or changes to frequency can be used to maintain a constant input impedance to the CPT link as the separation distance between the transmit and receive electrodes is varied. In previous works, frequency compensation required continuous adjustment to track spatial changes, thereby requiring broad operating bands for broad spatial ranges. Here, we combine discrete frequency compensation with continuous impedance compensation to keep transmission frequencies within specific narrowband ISM channels. This work presents a practical implementation of spatially robust CPT systems. We consider the limitations of band-limited wireless power channels and apply frequency compensation as discrete frequency changes that are switched between different bands. In this way, frequency compensation offers coarse impedance control for large spatial changes. We then combine the band switching with load impedance compensation methods that offer fine impedance control. Experimental results verify the design concepts using an electrode configuration that is compatible with a differential six-plate CPT system. The results include measurements of the compensated CPT link in isolation as well as an evaluation of compensation applied to an end-to-end system with a class-E amplifier and class-E rectifier. The results show that multi-band operation significantly extends the spatial robustness of the system. Compared to single-band operation, there is a 50\% increase in the compensation gap range in the dual-band system. The choice between multi-band and single-band systems depends on application requirements: multi-band adaptive load compensation works well in systems with large spatial variations, while single-band suffices for applications with modest spatial variation and defined range constraints.

A document by Jinzheng Zhao . Click on the document to view its contents.

Amazing…..? Negative and positive sum of prime number are equal…very interesting …please see this working so beautiful. New conjectures in number theory I think

An algorithm that solves a linear program with planes exterior to the feasible region is described. Given a linear program in a standard form, a constraint derived from the objective function is added to make the program infeasible, the objective function is removed. The added constraint is incrementally and unidirectionally moved toward the solution vertex in cycles. On reaching the solution vertex the program becomes feasible, this signals both the solution of the program and termination of the algorithm. Because of unidirectionality, the algorithm is guaranteed to terminate. With N variables, B-bit fixed point arithmetic, and feasibility testing complexity of O(f(N)), the complexity of the incremental exterior plane algorithm is 2B O(f(N)). With binary search, the complexity is ~O(B f(N)). Feasibility checking may be done by finding the optimum of a related LP in each cycle. Unlike in established LP methods it is only necessary to know if an intermediate constraint set is feasible or not, neither the optimum of the related LP nor the basic feasible point obtained is used or needed. The algorithm works efficiently with deformed products. Thus with a 43-dimensional Klee-Minty cube (which would require > 1012 pivots with the simplex method) the solution is obtained in less than 100 binary search cycles, similar results are obtained with Goldfarb-Sit cubes. The exterior plane algorithm is compared qualitatively with the simplex, ellipsoid, and interior point algorithms and miscellaneous others. Â

Robotic Prosthetic Hands (RPH) are made with the aim of assisting amputees with activities of daily living (ADL). Upper limb amputations lead to physical and mental difficulty because of the many uses of the human hand. Researchers have been working to improve sensory feedback and dexterous manipulation in these devices to improve embodiment and reduce the abandonment rates. Efforts to implement sensory feedback have been explored by researchers in the field of biomedical engineering whereas efforts to improve dexterous manipulation are directed towards the fields of mechanical engineering and computer science. There are numerous problems faced in the field of robotic prostheses such as end point control, sensory feedback, the weight/dexterity dilemma, and the implementation of invasive methods. RPH learning to grasp from touch with tactile sensors is an additional area of research that is promising for robotic grippers. The design of RPH requires multidisciplinary knowledge related to physiology, anatomy, electronics, mechanical design, and software which renders it a complex challenge. This body of work offers a holistic review of the state of the art of current RPH and the subsequent opportunities for advancement in the field.

A document by Akash Kadechkar . Click on the document to view its contents.

We investigate how the performance of a cascaded metasurface system can change when it is designed with physically realizable unit cells as opposed to unit cells based on impedance boundary conditions. In the provided example, we have observed that the performance of the physically realizable metasurface system has degraded significantly close to its edges.

Globally distributed microelectronic supply chains have disrupted trust in silicon hardware and have drawn academia’s attention toward different scenarios of malicious circuit modifications, i.e., hardware Trojans. This dynamic hardware environment, including open-source approaches and evermore outsourcing, requires constant reassessment of offensive and defensive aspects. Based on an untrusted foundry model, this work assesses the concrete technical realizations of layout-only modifications via design file editing, mask editing, or in-line alterations. Furthermore, the attack possibility on different modules within a system on a chip is qualitatively evaluated. Consequently, a modification is demonstrated on an SRAM-’PUF’ module. To link the attack point-of-view with a defensive measure, we propose a hardware reverse engineering-based countermeasure, which is non-reliant on a golden layout. Through a novel approach relying on inherent polygon properties, potentially occurring modifications are detected via clustering and a statistical evaluation of the intra-cluster distributions.  Finally, the approach is demonstrated on samples from 7 nm to 150 nm, for which a modification detection rate between 95% and 100% is reached for all evaluated samples.

This letter addresses the impact of the hardware impairments on the data transmission over Terahertz (THz) band.

Understanding and combatting biodiversity loss are critical tasks facing our planet. They are made especially difficult because much of the earth’s biodiversity is concentrated in abundant and speciesrich groups of invertebrates like insects. Traditionally, samples of insects have been analyzed manually by experts using morphology. Not only does this necessitate taxonomic expertise, but it is also error-prone, time-consuming, and often involves commercial microscopes that are too expensive for many countries in the Global South where most species are found. The alternative to expert sorting with morphology is the use of DNA barcoding. However, this requires a well-equipped laboratory and an entirely different skill set. We present an alternative solution: a low-cost, open-source photomicroscope for taking high-resolution, focus-stacked images that can be used for insect classification: the Entomoscope. We describe two different versions of the Entomoscope, a standalone version that can be operated without additional hardware and an even simpler Version, that requires a computer. We show that the optics are of sufficiently high quality to classify specimens with >95% accuracy into 15 different types of insects (mostly ’families’ according to the Linnean classification). The classifier can be successively extended or individually trained for specific classification tasks. Here, we provide building instructions, 3D files, and a list of commercially available parts so that everyone can build their own Entomoscope. Open-source DIY hardware like the Entomoscope facilitates affordable, cutting-edge biodiversity research by entomologists around the globe.

The main subject of this work is to make a detailed comparison between the performance of free-space optical (FSO) and terahertz (THz) links under different conditions.

The self-supervised training method can fully leverage the structural characteristics of unlabeled datasets, enhancing the technical prowess of natural language processing. We explore the application of self-supervision’s advantages to query expansion. To achieve this, we amassed a substantial query corpus and trained it using a generative model. Simultaneously, we introduced the SG-BERT/GPT2 extension method, leveraging two generators, BERT and GPT2. We also conducted an analysis of the impact on code search before and after generative model training, as well as the influence of the original query length on the expanded model. Experimental results consistently demonstrate the superiority of the trained model over the untrained one. Additionally, in deep learning-based code search methods, query expansion proves most effective for datasets with relatively short original query lengths, while longer expansions yield the opposite effect. Finally, when compared to various established query expansion methods, the SG-BERT/GPT2 expansion method, rooted in the self-supervised generative model, consistently outperforms them across multiple language test datasets. With ongoing self-supervised generation and continuous optimization of the query corpus, further enhancements in effectiveness can be achieved.

In this work, we investigate the performance of free space optical (FSO) communication systems based on modulating retroreflector (MRR) array.

In this article, our goal is to design a permanent long backhaul link using unmanned aerial vehicle (UAV) relays and charge stations (CSs) to transfer data from the nearest core network to disaster area.

This work presents a highly parameterised CGRA-based accelerator that we developed for an extensive Design Space Exploration activity on design parameters. The description starts from the CGRA building blocks, the Functional Units, and progresses towards the top level of the architecture, represented by the Node component, which is composed of an NxM matrix of Processing Elements. For each level of the hierarchy, we describe the HDL design parameters affecting the run-time reconfigurability of the accelerator, delving deeper into the functionality of the architecture. Outcomes are reported after synthesis on TSMC 40nm standard-cell technology.

A document by Yasmina Zaky . Click on the document to view its contents.